​
Overview

A voice is a saved profile you can reuse across generations, stories, and the API. In PeachDesk Voice Studio, voices come in two flavors that map to two different ways of getting a voice: Both types live in the same Voices tab and behave the same way at generation time — pick the type that matches your goal and follow the workflow below. The models that power both types run on PeachDesk’s GPU infrastructure, so there is nothing to download or install: you select an engine from a dropdown and Voice Studio does the rest.
Not sure which to use? Cloning gives you a specific voice but needs clean audio. A built-in voice gives you good results instantly, but you don’t get to choose who it sounds like.

​
Workflow A — Cloned Voices

Use this when you want to replicate a specific person’s voice from a recording.
1

Prepare audio

Gather 10–30 seconds of clear speech with minimal background noise.
2

Create the voice

Open the Voices tab and choose + New Voice → Clone from audio, then pick a cloning engine (VAANI, VAANI-ENGLISH, VAANI-MULTILINGUAL, or VAANI-INDIC — all four can clone). The engine runs on PeachDesk’s servers — no model download required.
3

Upload or record a sample

Drag in an audio file, or record directly with the in-browser recorder.
4

Generate to test

Use the new voice to generate a test phrase. If quality is poor, add more samples.

​
Audio requirements (cloning only)

Duration

10–30 secondsToo short: poor quality. Too long: unnecessary.

Clarity

Clear speechNo background noise. No music or overlapping voices.

Quality

High fidelity44.1 kHz or 48 kHz sample rate. Minimal compression.

Content

Natural speechConversational tone. Complete sentences.

​
File formats

Supported formats:
  • WAV (recommended) — lossless quality
  • MP3 — acceptable, use minimal compression
  • M4A — acceptable
  • FLAC — lossless alternative
Use WAV for best results. Avoid heavily compressed formats.

​
Recording tips

  • Record in a quiet room
  • Turn off fans, AC, and appliances
  • Close windows to reduce outside noise
  • Use soft furnishings to reduce echo
  • 6–12 inches from the mouth
  • Slight angle to reduce plosives (p, b, t)
  • Use a pop filter if available
  • Maintain a consistent distance
  • 44.1 kHz or 48 kHz sample rate
  • 16-bit or 24-bit depth
  • Mono is fine (stereo will be converted)
  • Avoid automatic gain control

​
Speaking style

  • Natural pace — don’t rush or speak too slowly
  • Clear articulation — pronounce words clearly
  • Consistent volume — maintain steady loudness
  • Normal tone — speak as you normally would
  • Complete sentences — avoid fragments or “ums”

​
Multiple samples

Adding multiple samples can significantly improve quality:

Robustness

The model learns a more complete representation.

Versatility

Handles different speaking styles better.

Quality

Reduces artifacts and improves naturalness.

Consistency

More reliable across different texts.
Consider adding samples with:
  1. Different tones — casual, formal, excited, calm
  2. Different content — narratives, questions, statements
  3. Different recording conditions — studio quality, room acoustics
All samples should be from the same speaker. Mixing voices will produce poor results.

​
Processing existing audio

If you have existing audio (podcasts, videos, and so on):
1

Find clean speech

Look for segments with just the target speaker — no background music and minimal noise.
2

Use an audio editor

Tools like Audacity or Adobe Audition: cut clean 10–30s segments, remove silence at the start and end, and normalize volume.
3

Export as WAV

Save as a high-quality WAV file, then upload it in the Voices tab.
For light background noise, use Audacity’s noise reduction with gentle settings — over-processing introduces artifacts.

​
Testing and iteration

After creating a cloned voice:
1

Generate a test

Try a simple phrase: "Hello, this is a test of my voice."
2

Evaluate quality

Listen for natural tone, clear pronunciation, proper prosody, and a lack of artifacts.
3

Iterate

If quality is poor: add more samples, try different source audio, or check sample quality.

​
Common issues

Cause: poor quality samples, or samples that are too short.Fix: use longer, higher-quality samples.
Cause: the sample tone doesn’t match the desired output.Fix: record samples in the style you want to generate.
Cause: background noise or audio issues in the samples.Fix: clean up the samples, or re-record in a quieter environment.

​
Workflow B — Built-in Voices

Use this when you want a ready-made voice without recording anything. One engine ships built-in speakers: VAANI-INDIC (16 voices — a female and a male speaker for each of eight Indian languages). It is hosted on PeachDesk’s infrastructure — just pick a voice and start generating.
1

Create the voice

Open the Voices tab and choose + New Voice → Built-in voice, then select VAANI-INDIC as the engine.
2

Pick a voice

The engine’s voice catalog appears. Click any voice to preview it.
3

Name and save

Give the voice a name. No audio sample is required.
4

Generate

The voice is ready immediately — use it from the Generation tab.
Built-in voices are locked to their source engine. If you switch to a different engine in the Generation tab, the voice is greyed out because it only exists in that engine. Clicking a greyed-out voice switches the engine back automatically.

​
Shaping a built-in voice

A built-in speaker is fixed, but the delivery is not. With a VAANI-INDIC voice selected on the Generation tab, open the Fine-tune panel to adjust expressiveness, pacing (CFG), sampling temperature, and repetition penalty. For pitch, speed, and other post-processing, use the Effects tab — those are applied to the rendered audio, so they work with every engine.

​
Advanced tips

​
Celebrity and character voices (cloning)

For cloning public figures or characters:
  1. Legal considerations — make sure you have the rights, or that it is clearly fair use.
  2. Source quality — find high-quality interview audio or clean clips.
  3. Consistency — use clips where they speak similarly.
  4. Multiple samples — very important for recognizable voices.

​
Accent and dialect (cloning)

Cloning models preserve accent and dialect:
  • British English samples generate British English output.
  • Southern-accent samples produce Southern-accent output.
  • Regional pronunciations are maintained.

​
Emotion transfer (cloning)

The emotional tone of the samples affects generation:
  • Energetic samples → energetic output
  • Calm samples → calm output
  • Mix samples for a more versatile voice
For built-in voices there are no samples to vary, so use the Fine-tune panel’s expressiveness control to dial the delivery up or down instead.

​
Managing voices

​
Organization

  • Descriptive names — for example, “John Smith – Professional Narrator”
  • Add descriptions — note recording conditions, use cases, or which built-in voice
  • Language tags — mark the primary language
  • Archive unused — keep the voice list manageable

​
Sharing within your workspace

  • Voices you create are available to your PeachDesk workspace.
  • Cloned voices are stored as voice embeddings rather than the original audio.
  • Built-in voices are stored as engine + voice ID metadata only, and play back instantly for anyone in the workspace since the engine runs on PeachDesk’s servers.

​
Next steps

Generate Speech

Use your voice to generate speech in the Generation tab.

Build Stories

Create multi-voice narratives in the Stories tab.

Apply Effects

Apply versioned effect takes to your generations.